Relating instance hardness to classification performance in a dataset: a visual approach

Machine Learning - Machine Learning studies often involve a series of computational experiments in which the predictive performance of multiple models are compared across one or more datasets. The...     

About the origins of residual stresses in <Emphasis Type="Italic">in situ</Emphasis> consolidated thermoplastic composite rings

In a context of cost reduction, in situ filament winding of thermoplastic matrix composites becomes an appealing process. As residual stresses could considerably affect the produced part, models were proposed to predict process-induced residual stresses. After developing a validated thermal model of the process, mainly three different aspects are here addressed: the continuous bonding occurring during the process, the effect of the processed layer on the structure, and the effect of the curvature of the mandrel. While stresses coming from the continuous bonding appeared to be negligible, consequent levels of stresses can be achieved due to an iterative compression of the structure by the tow (supposed to be under tension). The mandrel properties and the tow tension profile followed during winding are essential parameters that might induce several different stress states. A comparison between measured and computed end-to-end openings of split rings illustrates the accuracy of the proposed models.     

Solving identity delegation problem in the e-government environment

At present, many countries allow citizens or entities to interact with the government outside the telematic environment through a legal representative who is granted powers of representation. However, if the interaction takes place through the Internet, only primitive mechanisms of representation are available, and these are mainly based on non-dynamic offline processes that do not enable quick and easy identity delegation. This paper proposes a system of dynamic delegation of identity between two generic entities that can solve the problem of delegated access to the telematic services provided by public authorities. The solution herein is based on the generation of a delegation token created from a proxy certificate that allows the delegating entity to delegate identity to another on the basis of a subset of its attributes as delegator, while also establishing in the delegation token itself restrictions on the services accessible to the delegated entity and the validity period of delegation. Further, the paper presents the mechanisms needed to either revoke a delegation token or to check whether a delegation token has been revoked. Implications for theory and practice and suggestions for future research are discussed.     

Consequences and conjectures in preordered sets

In a preordered set, or preset, consequence operators in the sense of Tarski, defined on families of subsets, are introduced. From them, the corresponding sets of conjectures, hypotheses, speculations and refutations are considered, studying the relationships between these sets and those previously defined on ortholattices. All the concepts introduced are illustrated with three particular consequence operators, whose behavior is studied in detail. The results obtained are applied to the case of fuzzy sets endowed with the usual pointwise ordering.     

Addressing data complexity for imbalanced data sets: analysis of SMOTE-based oversampling and evolutionary undersampling

In the classification framework there are problems in which the number of examples per class is not equitably distributed, formerly known as imbalanced data sets. This situation is a handicap when trying to identify the minority classes, as the learning algorithms are not usually adapted to such characteristics. An usual approach to deal with the problem of imbalanced data sets is the use of a preprocessing step. In this paper we analyze the usefulness of the data complexity measures in order to evaluate the behavior of undersampling and oversampling methods. Two classical learning methods, C4.5 and PART, are considered over a wide range of imbalanced data sets built from real data. Specifically, oversampling techniques and an evolutionary undersampling one have been selected for the study. We extract behavior patterns from the results in the data complexity space defined by the measures, coding them as intervals. Then, we derive rules from the intervals that describe both good or bad behaviors of C4.5 and PART for the different preprocessing approaches, thus obtaining a complete characterization of the data sets and the differences between the oversampling and undersampling results.     

A particular approach for the Three-dimensional Packing Problem with additional constraints

This paper focuses on a restrained concretization of the general NP-hard Container Loading Problem that arises from a real world application. This particular problem can be informally described as: given different sets of bins and boxes, find the packing of the boxes into the smallest number of bins obeying some additional restrictions on the placement. The mathematical programming formulation that appears to better model this application is the Three-dimensional Bin-Packing Problem (3D-BPP) which is no more than an extension of the classic Bin-Packing Problem to the orthogonal packing of solid objects.We present an alternative non-linear formulation for the 3D-BPP version as well as new designed heuristic algorithms (CPBOX and BOXCP) for the approximation of the optimal solution using boxes and bins of heterogeneous dimensions. Moreover, we also present a heuristic specifically designed to work under very particular storage application restrictions. The numerical results indicate that all of these new heuristics have good performance standards especially when applied to replications of real input data.     

An experiment for improving students performance in secondary and tertiary education by means of m-learning auto-assessment

Mobile learning is considered an evolution of e-learning that embraces the ubiquitous nature of current computational systems in order to improve teaching and learning. Within this context it is possible to develop mobile applications oriented to learning, but it is also important to assess to what extent such applications actually work. In this paper we present a new tool designed to reinforce students’ knowledge by means of self-assessment. Improvement in student achievement was evaluated and an attitudinal survey was also carried out to measure student attitudes towards this new tool. Three different experimental groups were selected for this research, with students aged from 14 to 21 years old, including high-school and university students. Results show that this kind of tool improves student achievement, especially amongst younger learners, with a relatively low impact on current teaching activities and methodology.     

Context-sensitive dependency pairs

Termination is one of the most interesting problems when dealing with context-sensitive rewrite systems. Although a good number of techniques for proving termination of context-sensitive rewriting (CSR) have been proposed so far, the adaptation to CSR of the dependency pair approach, one of the most powerful techniques for proving termination of rewriting, took some time and was possible only after introducing some new notions like collapsing dependency pairs, which are specific for CSR. In this paper, we develop the notion of context-sensitive dependency pair (CSDP) and show how to use CSDPs in proofs of termination of CSR. The implementation and practical use of the developed techniques yield a novel and powerful framework which improves the current state-of-the-art of methods for automatically proving termination of CSR.     

Control of absorption columns in the bioethanol process: Influence of measurement uncertainties

The alcohol lost by evaporation during the bioethanol fermentation process may be collected and recovered using an absorption column. This equipment is also used in the carbonic gas treatment, a by-product from the sugar cane fermentation. In the present work, the development of nonlinear feedforward–feedback controllers, based on neural network inverse models, was proposed and tested to manipulate the absorbent flow rates. The control purposes are: to keep low ethanol concentration in the effluent gas phase from the first absorption column (ethanol recovery column); and to reduce the residual water concentration in the CO₂ gas effluent from the second tower (CO₂ treatment column).Based on simulation studies, the neural network (ANN) controller performance was compared with the conventional PID control scheme application. The best ANN architecture was set up according to the Foresse and Hagan (1997) criterion, while the PID parameters were found from the well-known Cohen–Coon Equations and trial-and-error fine tuning.Initially, performances were evaluated for the system without concentration measurement uncertainties. From these tests, the ANN controller presented the smallest response time and overshoot for regulator and servo problems. Three uncertainty levels were applied afterwards: 5%, 10%, and 15%.The ANN controller outperformed the PID for all uncertainty levels tested for the ethanol recovery column. For the CO₂ treatment column, the ANN controller proceeded successfully under uncertainties of 5% and 10%, while the PID did not deal properly with uncertainties above 5%. The statistical F-test, besides the ITAE, ISE, and IAE performance criteria, were calculated for both controllers applications and then compared. They proved the superiority of the ANN control scheme.Using appropriately the proposed well-controlled absorption columns increases the efficiency of the bioethanol production plant and can also provide carbon credits by avoiding CO₂ emission into the atmosphere.